About . . . . . . Classes . . . . . . Books . . . . . . Vita . . . . . . . Links. . . . . . Blog

by Peter Moskos

December 23, 2016

No, it's not just Chicago

Homicide is up at a record setting two-year pace. But you wouldn't know it from yet another press release by the Brennan Center. I think they time these to provide a "crime is not a problem" narrative to journalists, quite a few of whom are about to write year-end stories with the headline: "Oh shit, homicide is way up!" While I can't really question the motives of crime increase deniers, I can debunk the worst of their claims:

False claim #1) "Nationally," says the Brennan Center, "The murder rate is projected to increase 31.5 percent from 2014 to 2016 — with half of additional murders attributable to Baltimore, Chicago, and Houston."

This is so not true, I don't know where to start.

It's the "half" part I'm talking about. (In a previous post I mentioned that 31.5 percent should be 23.2 percent.)

Collectively, Baltimore, Chicago and Houston will see about 540 more murders in 2016 than in 2014 (my numbers are 1,406 vs. 866). Meanwhile, nationally, there will be roughly 3,600 more murders (17,768 vs. 14,164). Ergo, QED, and I told you so: Baltimore, Chicago and Houston account for 8 percent of all murders and fifteen percent of the additional murders in the US from 2014 to 2016.

8% = 1,406/17,768
15% = (1,406-866)/(17,768-14,164)

A few days ago, they doubled down, "The 2016 murder rate is projected to be 14 percent higher than last year in the 30 largest cities. Chicago is projected to account for 43.7 percent of the total increase in murders." I guess this depends on what "total" means. Because Chicago will be 14 percent of the "total" national increase.

14% = 300 / (17,778 - 15,696)

Now I think Chicago might be 43.7 percent of the increase in the top 30 cities. But that is some meaningless made-up playoff stat. I mean, if you look at the top three cities, it turns out Chicago makes up 102(?) percent of the "total increase." When your formulas can get you absurd results, it means you're doing it wrong!

Are they lying or just in error? Are they making honest mistakes or intentionally misleading? I don't know. But these "fact" are up there, cited by many, corrected by none.

False claim #2) It's all Chicago's fault

Stop blaming Chicago just because it's leading the pack.

Imagine I said, "the Chicago Bulls of the 1990s were really good!" And you replied, "No, not really, because Michael Jordan was playing for them."

Or if I said, "the Alps are a really tall mountain range!" And you replied, "Well, they wouldn't be so tall without Mt. Blanc."

It's difficult to respond to the reply because though substantively irrelevant, it is technically and semantically correct. (Is there some rhetorical term for this kind of distraction argument? Something in Ancient Greek for "hey look, a squirrel!") Indeed, Michael Jordan led the Bulls and the Alps would not be as tall without its tallest mountain. But so f*cking what? The rest of the Bulls were good basketball players. And the Alps would still be tall without its tallest mountain!

The national increase in homicide is a problem with or without Chicago. As I've written before, you can remove Chicago and other cities (not that you should) and the increase in homicide is still very large (albeit, yeah, smaller). There will be about 17,800 murders in 2016. About 4 percent of these [800 / 17,800] happen in Chicago. The two-year increase in Chicago homicide is about 10 percent of the national total  [(777-407) / (17,800-14,164)]. Conversely, 90 percent of the national increase in homicide is not in Chicago.

Chicago, of course, is special. But let's not get distracted. The rise in homicide is not just a problem in "a few, select cities." It's a problem except in few select cities.

False claim #3) Some years murder goes up and some it goes down.

Well, yes, that's true. But not right now. The increase in murder is not, despite what they say, some story of random statistical year-to-year fluctuation:
A similar phenomenon occurred in 2015, when a group of three cities — Baltimore, Chicago, and Washington, D.C. — accounted for more than half of the increase in murders. This year Baltimore and Washington, D.C., are projected to see their murder rates decline, by 6 percent and 18.6 percent, respectively.
Well, you know, up one year and down the next.... Except murder is up. Not up-and-down. And when a few cities go up one year, and other go up next, and all of them go up overall, it's we who took advanced statistics call "a trend."

And dammit, cause Baltimore is personal, Baltimore is not a city running contrary to this upward trend. 2015 was off-the-charts bad for crime Baltimore, after the riots. 2016 was horrible, as in the worst year ever... except compared to 2015. I mean, it's good 2016 is not worse than 2015, but it's not good. To say crime is down in Baltimore is disingenuous at best. The highest murder rate ever -- 2015 -- should not be the new normal by which we judge success.

False claim #4) "Concerns about a national crime wave are still premature."

OK, but if not now, when? 27(?) of the 30 biggest cities have seen an increase in murder over the past two years. We need to do more than simply, "suggest a need to understand how and why murder is increasing in some cities." Are we not supposed to care as long as violence remains segregated in poor segregated black neighborhoods? And we know why murder is increasing in some cities. It's not rocket science. "Lack of socioeconomic opportunity has long been credited with high levels of crime." Yes, no shit. Of course violence happens more in shitty neighborhoods without "socioeconomic opportunity." But that's neither here nor there because "socioeconomic opportunity" hasn't gone down the shitter in the past two years. "Socioeconomic opportunity" might explain (part of) the problem. But it doesn't explain the increase in homicide in the past two years. (Last year, in fact, saw a record decrease in poverty).

False claim #5) Sure homicide is up, but not crime overall.

First of all, if you think rising homicide doesn't matter because other crime is steady, for shame. Second, homicide matters more than other crime. Period. If homicide is up, stop right there. But the statistical problem is that data on crime overall is not that good. A lot of crime isn't reported. We know that. (There's the NCVS, but they have their own problems.)

And I'm not even talking about intentional data manipulation here. An unreported mugging is still a mugging. And the reality gap between crime and reported crime is worse than you think. And it's even worse from a statistical perspective, because there's no reason to think that errors and missing data are consistent year-to-year. (I'm a big stickler about non-random missing data, if that means anything to you.)

When police get out of their car less, they make fewer arrests. And when cops make fewer arrests, *poof* reported crime goes down in sync. It's like a crime never happened. (Except, of course, it did.) If cops get out of their cars more, if cops confront more criminals, there will be more crime recorded. (Which can be falsely interpreted as an increase in crime.)

So when it comes to crime, I trust murder. And very little else. Conveniently, homicides are correlated with all kinds of crime. So if homicides are up (and I'm using "homicide" as synonymous with "murder"), and somebody tries to tell you crime is steady... you shouldn't believe them (even if they believe what they're saying). Question crime data. Hell, question all data. How else will we know if data is valid or from some fakenews meme. And when bad data gets out there, it's a problem for all data.

False claim #6) Crime is still at a historical low

Kind of, sort of. But who's to say what "normal" is? Why should the high crime decades be the standard bearer? Why not the lower crime decades? Crime is kind of where it's always been, if one excludes the high crime 1970s and 1980s. And certainly by civilized world standards we're still crazy murderous.

The point shouldn't be some arbitrary historical date but current trends. And we don't apply that "historically low" BS to other issues. You know what other things, big picture, are at historical lows despite recent uptick? Racism, hate crimes, authoritarian rulers, scurvey, and the bubonic plague. I can't put this strongly enough, but f*ck historic lows (and keep in mind when it comes to crime our "lows" are pretty high). When bad things rear their ugly head, of course we worry and try to nip the problems in the bud before they become an epidemic.

21 comments:

Jay said...

As I understand it, you are saying: Murder is up, therefore other crimes must be up. The reason the statistics do not show that increase is that cops aren’t getting out of their cars.

I assume that you’re talking about Index crimes. I thought that most of these crimes got on the books from a sequence of victims calling the cops, and cops recording the crime. Does either of these involve cops not getting out of their cars? If I call the cops to report that I was mugged or that my house was burglarized or that my car was stolen, does the Ferguson effect deter cops from recording the information?

I'm too lazy to look up this stuff, but is the NCVS data for 2015 available? Your explanation would suggest that if crime is really up and that UCR stats don't show that increase, then the gap between NCVS and UCR would be greater in 2015 and 2016 than in previous years. Or that the NCVS would show an increase in robbery, burglary, auto theft while the UCR shows them unchanged. Is that the case?

Peter Moskos said...

If you call cops and speak to cops, they will record the crime (generally).

But there's the "on-view" crime, where you get out of your car to investigate a few people standing around talking. Your spidey sense is tingling. (Indeed, you don't even have reasonable suspicion to make a stop.) But you're a good cop and you ask through your window, "what's up?" One person is distraught and you can't tell if she is hurt so you put your car in park to get out and see and interact better. Then (much to your regret) you go down a rabbit hole of other people's personal misfortune, bad luck, and bad choices. Just when you're wondering why you didn't stay in your car and keep driving, you realize you've got a life story, a history, a crime, and a victim; perhaps even a suspect, probable cause, and (perhaps even) an arrest. Suddenly *bam* you've got paperwork and a recorded index crime.

But in addition to that on-view scenario, if a victim doubts police legitimacy, they're less likely to the call the cops ("cops take too long, they're mean, they won't help me, they might shoot me, they're racist" etc.). Presumably that has gone up as well. But how much is anybody's guess.

NCVS did *not* show an increase, up to 2015. Here's their most recent summary:

https://www.bjs.gov/content/pub/pdf/cv15.pdf (open this link for discussion, below).

Note that with their confidence interval, it's hard to see any change any year, at least going back to 1999.

But I've been getting more and more suspicious of the NCVS. (see the link on my post http://www.copinthehood.com/2013/11/crime-is-up-no-wait-crime-is-down.html)

I think the NCVC has major polling problems related to society and their methods. Compounded with reliabilty issues inherent in any victim-reported account. The NCVS survey claims an 86% response rate. 86% of potential respondents were successfully contacted and gave accurate crime data? That really stretches the imagination. Especially in high-crime and hidden communities.

Also, 163,000 people is a lot, but is that enough of a sample? I don't know. But if we're concerned with poor black high-school-drop-out males, for instance, we're down to maybe 1-2 percent of the population, presumably with a disproportionately higher non-response rate.

Look at NCVS Figure Two on page three (link you opened above). There's simply no way in hell -- given that gun homicide was up so much in 2015 -- that nonfatal firearm victimization was down in 2015. No way. The NCVS must be wrong here. Certainly not with year-to-year trends. Certainly not with violence. Certainly not with a population that won't admit to being shot to police, hospitals, or a NCVS/census worker knocking on the door. I mean, think of the guys Alice Goffman wrote about. You think they're responding to the NCVS survey? But if a survey fails to capture those most at risk of violence, how can it measure violence? And no, "weighted samples" based on race and age are not enough.

Figure 4 on the NCVS is also interesting. But again, how can measure non-reporting of crime if you've got a non-random issue of non-respondents? "The rate of violence and serious violent victimization committed against males decreased from 2014 to 2015." That is the statement I do not believe, given the increase in homicide.

I'd be interested in looking at CDC data, but I've never gone there. But one could do worse than starting here: https://www.cdc.gov/nchs/data/nvsr/nvsr65/nvsr65_04.pdf#tab18

Peter Moskos said...

Look at Appendix Table 4 on page 19.

https://www.bjs.gov/content/pub/pdf/cv15.pdf

In terms of numbers, the NCVS shows that nonfatal firearm violence from 2014-2015 *deceased* from 466,000 incidents to 285,000 incidents. Fat chance. This is crap data. And the check out the 95% confidence interval range. It's large. Too large. So potentially, there could have been a 1-year 80% decrease in gunshot victims in 2015. Or a 21% increase. Take your pick.

Jay Livingston said...

I went to the UCR for the Chicago numbers – 2012-2015 (the most recent year available).
Murder in 2015 was up 16% over 2014.
Aggravated Assault up 4-5%.
MVT a small increase – 10,222 vs. 10,023 – but far below the 17,000 of 2012.
Robbery – small decreases from 2014 (1%) but big decrease over 2012 (28%)
Burglary – 9.5% decrease from 2014; 42% decrease over 2012.

Suppose these other crimes had been going up in 2015 at the same rate as murder (16%). Here are the numbers predicted and the numbers in the UCR
Robbery – 11,372 - 9,649 (difference - 1,724)
Burglary - 16,863 - 13,151 (difference - 3,712)
MVT - 11,626 - 10,222 (1,405)

How many of those 3700 presumably unrecorded burglaries were from a reduction in on-view? How many from reluctance to call the cops? How many of the 1700 robberies? How many of the 1,400 MVTs?

Tombstone courage said...

The crime statistics are not worth the paper they are printed on. Homicide stats are the closest to being valid, because you need a head to go with it. But even then, it can be buried as a death investigation. A lot of crime is unreported, so the premise that you can even establish a baseline for monthly or yearly comparison is false.

Liberaltarian . . . said...

It seems odd to me that you rely on the numbers when they support your position (increase in homicide) but disparage the numbers when they don't (increase in overall crime). You don't even mention what the Crime Victimization Survey shows regarding overall crime (if I'm not mistaken, it, like the UCR, shows overall crime basically flat). You're right to castigate the Brennan Center for playing fast and loose with the numbers when they run counter to their assumptions. But, labeling the statement "Sure homicide is up, but not crime overall" as false when the stats from both the UCR and NCVS show that this statement is true, puts you in the same position as the Brennan Center as to playing fast and loose with the numbers.

Peter Moskos said...

I don't rely only on numbers that I agree with. I rely on crime numbers only when they are homicide data. That's it! (or shootings, if accurately counted) If "crime data" were up while homicide numbers were down, I'd be the first to herald the very real drop in crime.

In fact, I don't think you can find me ever saying anything about "crime rate." I don't generally look even look at it. (Though sometimes I do write the word "crime" or "violence" when I'm really basing my statement only on homicide.) I've always only trusted homicide, at least in the US where enough people kill each other regularly. (For other countries I sometimes use "crime rate," by necessity. But data collection in these other countries tends to be better and nationalized.)

And I did mention the NCVS in a comment above, even with *emphasis*. But I've never relied on the NCVS, and I'm not going to start now. I didn't believe the NCVS 4 years ago when they described a huge rise in crime. I don't believe them now. Also, historically, the NCVS underestimates murder (that's a methodology joke, by the way).

Now I may not convince other people that the link between homicide and violent crime is ironclad and better than the data on violent crime. But that's OK. Because I don't really care if crime is *not* up. That's just a distraction to me.

Homicide is up. And that trumps all. I say the two are inherently linked, but it doesn't matter if I'm wrong and they're not. Perhaps somehow homicides are up 25 percent and robbery and other forms of assault are not up at all. If others want to say more murder isn't a problem *because* other crime isn't up overall, everywhere, and at a consistent rate of increase, well I say shame on them.

Also, according to the UCR, crime actually *is* up. It's just not up a lot. But when others say "crime isn't up," well, technically, that's not true. But I don't want to quibble whether 3 percent (or whatever it is) is statistically significant or not.

Adam said...

I agree that proactive policing is probably down, and thus there are fewer crimes being reported "on view." But isn't another cause (perhaps a bigger cause) that lots of people who otherwise would have reported crimes have, in the last couple years, decided not to call the cops? Wasn't that the finding of that recent Papachristos, et al. study?
So lots of crimes that the cops wouldn't have discovered on view anyway (had proactive policing not declined) are not being reported.

As for the "Hey look, a squirrel!" question, maybe this list of fallacies will help. "Ignoratio elenchi," perhaps?

Otis Blue said...

What do you expect from a group of people ("...some of the most knowledgeable observers of criminal justice today: journalists, academics, justice practitioners, think-tankers and advocates" -- note the absence of police officers) that observed the homicide trends of 2016 and overwhelmingly (80%) feel that police shootings of civilians is this country's biggest criminal justice challenge. These are the people that control the narrative and the agenda; they write the articles, appear on the panels, and advise our leaders.

http://thecrimereport.org/2016/12/22/a-troubled-and-troubling-year-in-criminal-justice/

I think it boils down to NYC. The historically low and still decreasing murder rate there provides the cover for the downplaying of the increasing murder rates elsewhere.

Thorn said...

Meanwhile 60 minutes ran a story on Chicago that basically blamed the police for the murder rate by no longer doing all the stops that the ALCU and media has criticized. Damned if you do...

Cory Giles said...

I do statistics in another area (biomedicine). So I'm curious about the methods and assumptions you use here in this field.

Obviously homicide is a worse crime than others. It is nicely unambiguous and probably has a high rate of being reported. I guess the only thing I would worry about is that homicide is a very rare event. 17K in a country of 300M is very rare, and measuring differences in rates of very rare events is tricky and prone to issues. I am curious if people model homicide with Poisson, hypergeometric, or something else?

All these claims (by you and Brennan Center) are in terms of percent change with no clue given as to what percentage change would roughly correspond with a significant increase/decrease. IMO, responsible researchers should not say "crime X is up" unless they mean "crime X is significantly up over last year", ideally controlling for SES etc. I guess it probably is, but we humble readers don't know.

Anyway, if the data from nonhomicide crimes does correlate *highly* with homicides (you say it correlates, but not how much), and the data are reasonably equally reliable in both types (you dispute this), then it might make more sense to infer homicides from nonhomicide crimes rather than the reverse because the amount of data would be larger for nonhomicide crimes. I would be very interested to know how strong that correlation is, and also, I would guess that whatever the number, various nonhomicide crimes are more correlated with each other than any of them are with homicide, because homicide is a very special kind of crime. That would suggest it might make sense to treat them differently.

It seems that the statements "I trust murder, and very little else" and "conveniently, homicides are correlated with all kinds of crime" are almost contradictory. If they are highly correlated, you can measure either one and get the same answer. If they aren't, then it suggests they should be treated as separate categories.

But as to your overall point, yes, all these claims by the Brennan Center seem absurd if you and they are using similar datasets and assumptions. Cherrypicking is especially bad. But OTOH, it seems to me that if you are interested in crime and not just homicide, you should use all the data available and model BOTH types.

Cory Giles said...

Also, why are these crime statistics always reported and analyzed on an annual basis? It seems to me you could draw more accurate conclusions without waiting until the end of a year if you did rolling/time-series analysis? Seeing things like "2016 (projected)" makes me cringe.

At the very least, if you are analyzing things on, say, Aug 1, then do the analysis per-year with annual cutoffs at Aug 1? Then you could say, as of Aug 1, "crime has significantly increased over last year compared to the year before" without using any projections, which probably diminish the credibility of the analysis in the public's eye. And then you are comparing apples to apples by having the same amount of time in the 2016 analysis compared to previous years.

I would like to see a plot that shows a continuous estimate of homicide rate or whatever on each day over time. It would have to be smoothed, obviously. Such a plot would be very useful if you were trying to figure out the causes. (I'm not suggesting you do this, I'm just remarking that it's odd to me that it isn't already the standard. I guess a nontechnical audience might have trouble understanding it).

Peter Moskos said...

Cory, thanks for your comments. They're good. And you're right. And generally good statisticians (which I am not, in the quant academic sense) do what you're saying. Steve Morgan at Johns Hopkins has done what you suggest in terms of rolling time-series analysis. (Poisson regressions were all the rage when I was in grad school. I never really understood them. From what I hear, they're out of fashion. Why? I do not know.)

But here's the thing: once advanced stat methods come into play, you lose 99% of the *educated* public. Very few people, even very few academics, can honestly judge and understand advanced statistical terms and concepts. At my core I am not a quantitative guy. That being said, I do like hard data. But here I'm not parsing side effects in a medical trial. I'm not even doing multivariate anything. My only real point is that homicides are up (I think crime is too, but that is a bit of a distraction). It's pretty basic. And yet people are still in doubt that it is and it matters. So I’m sticking with “25% in two years.”

I guess it comes down the audience and the purpose. I always try and write to an educated cop, one who understands numbers and basic statistical concepts. And English. But more fundamentally I’m not sure what your methods (which are better) would really add to that conversation. I’m arguing that crime data (homicide excepted) is seriously flawed. Advanced methods don’t and can’t compensate for that.

To answer a few other points. Why are crime data on a yearly basis? I don’t know. And back before computers, they used to come out quarterly (or 3 times a year?). Prof Richard Rosenfield has made that point repeatedly. So people like me spend time tabulated Baltimore homicides day by day (literally) to compare year-to-year data at specific times.

As to the n of homicide, 17,000 is plenty, even if a rare event, based on total population. And what always amazes me about homicide is that they *are* so consistent. One would think it’s not a natural function. And yet a city with 364 homicides a year *will* see an average (seasonally adjusted) of 7 per week. It’s amazing. But there are never some weeks with zero and others with 14. It’s almost like a water clock. I mean, statistically, that’s not surprising. But homicides aren’t statistics. It’s one person pulling a trigger and shooting another. But I’ll be damned if it doesn’t happen like clockwork.

And since homicide (and other crime) is so seasonally varied, it’s just easier to take the whole year and ignore the differences between September and January. I’m not saying this is best statistical practice. But it is more simple. And it avoids dishonest data manipulation.

As to the correlation between homicide and crime, well, if one looks at UCR data, the correlation is damn well perfect. But (I’m not looking at the data) if one argues that crime data is flawed, well, the whole thing is kind of futile, right? But generally, year to year, I believed the errors is crime data were consistent. What I’m saying now, in the past two years, is the missing data has changed. It is no longer consistent or random in the same way. Does that make sense?

Cory Giles said...
This comment has been removed by the author.
Cory Giles said...

> And since homicide (and other crime) is so seasonally varied, it’s just easier to take the whole year and ignore the differences between September and January.

OK, that makes sense. Especially when looking at long-term trends. It's just when you start making claims based on projected, partial data from one year compared to complete data for previous years that these sorts of corrections begin to possibly make a real difference in the conclusion. (I believe your conclusion, by the way, I'm just thinking partially in "reviewer-mode" or thinking how Brennan might respond here if they had competent statisticians).

> As to the n of homicide...what always amazes me about homicide is that they *are* so consistent.

(Until this year :) ) Poisson processes are amazing, aren't they? I only meant that the N is smaller than nonhomicides, not that the N was insufficient to draw conclusions about homicide.

> What I’m saying now, in the past two years, is the missing data has changed. It is no longer consistent or random in the same way. Does that make sense?

Sure, that would be a good reason to be skeptical of nonhomicide data. I just didn't see an explanation of why you think that in your post, except for a link to big differences between UCR and NCVS. It seems that this has become an issue, although you don't say it explicitly, because: A) historically homicides and nonhomicides have been highly correlated, but B) now homicides appear to be going up but not nonhomicides. Therefore, either something has happened to unlink homicides and nonhomicides, or one of the datasets is wrong.

And you are going with the latter. OK, as long as it can be demonstrated. OTOH, if one of the two among UCR and NCVS is more highly correlated with exogenous variables like homicide it might be a clue as to which one is more reliable.

> I guess it comes down the audience and the purpose.

Sure, this is a blog, not a paper. The reason I am thinking in terms of multivariate regressions and more finely-grained analyses is that, if you convince someone that homicide is indeed up, their next two questions will be:

1. When did it start going up? (necessitating a more finely-grained time-series analysis)
2. Why is it up? (necessitating multivariate regression)

It seems to me a little dangerous to simply present "homicides are up" to the public even if true, without also presenting data-driven reasons why it might be up. Otherwise all the usual suspects will draw their own biased conclusions: liberals will say guns and poverty, and conservatives will say it's cops being prevented from doing their jobs, immigration, and general moral decay of society (caricatures, obviously). Maybe you wouldn't present regression coefficients on a blog, but I am sure I'm not the only reader very interested in why this change, but also skeptical of any proposed answer that doesn't involve data.

Perhaps I was just thinking ahead a little, reading your post, asking how we might determine why it is up. I have read many of your posts but not all of them, so maybe you have addressed this question with data somewhere and I haven't seen it.

Otis Blue said...

@Cory,

The concept you appear to be struggling with here represents the gulf of understanding between data analyzers and data producers (note I didn't say data gatherers). I have a science background where I developed proficient statistical skills and was both data producer and analyzer. I then went into the military where I served as an intelligence analyst. I remember being frustrated with the incomplete nature of the data coming from my "sources" and even greater frustration with my "consumers'" apparent lack of comprehension of what was needed and possible from what I was getting. Basically, neither of us had any clue as to what the other needed or why. Now, I'm a cop, and I work the street pushing a patrol car around.

I had shared some anecdotes, but, meh, they weren't that exciting or unique. Instead I'll just summarize:

1. Property crimes aren't reported unless the insurance company requires it (meaning no insurance, no report) because it's too big a hassle to do so for most people.

2. Person crimes aren't reported unless the person believes police can help or an ambulance is required -- and even then I joke you aren't salty as a cop until you've cleared at least one attempt-murder call with "No Report".

Peter Moskos said...

I stumbled across this, for those who need a more academic presentation to be convinced.

How longer response time deflates crime stats:

https://nolacrimenews.com/2016/07/11/how-longer-dispatch-times-artificially-deflate-crime/

Melinda Waid said...

I tend to agree that crime data is unreliable. There are to many variations and unknowns to base fact upon. We can try and focus on current trends and try to solve those, but using past data to quantify current situations seems rather irresponsible to me. Population growth, changes in laws are just two things that make comparing past to present bias. One issue with reported violent crime is usually only the most severe is put in the log. Say someone murdered someone during a burglary or robbery, when the plea is done the robbery/burglary will be dropped because it is less important than getting a murderer off the streets. That does not mean the burglary/robbery didn't take place just that it didn't make the statistics file. There are lots of reasons that crime goes unreported and polls that ask citizens if they have been a victim of crime are unreliable because the average citizen does not know how the law is written, and some people unfortunately like the attention, and claim things that do not happen. Keeping track of homicide rates seem to be a reasonable way to track the progress or lack there of, of society but you cannot just use homicide rates you have to look at causation as well to attempt to get the full picture.

czrpb said...

can someone provide a link into the UCR? i didn't find data after 2014.

https://www.ucrdatatool.gov/Search/Crime/State/RunCrimeTrendsInOneVar.cfm

2010 14,722
2011 14,661
2012 14,866
2013 14,319
2014 14,249

Peter Moskos said...

Latest release: https://www.fbi.gov/news/pressrel/press-releases/fbi-releases-preliminary-semiannual-crime-statistics-for-2016

2015 data: https://ucr.fbi.gov/crime-in-the-u.s/2015/crime-in-the-u.s.-2015/resource-pages/downloads/download-printable-files

General: https://www.icpsr.umich.edu/icpsrweb/ICPSR/series/57

czrpb said...

thx!!